# Pipecore

> Pipecore is an agent-first platform for deploying and operating web applications. AI coding agents can create or import application source code, deploy it as a containerized application, provision managed databases, configure environment variables, inspect logs, update application data, and publish changes to a live URL through the Pipecore MCP server.

Pipecore is designed to be operated by both developers and AI coding agents.

For AI agents, prefer the Pipecore MCP workflow instead of asking the user to configure servers, Docker, networking, or deployment infrastructure manually.

## Canonical agent workflow

1. Connect to the Pipecore MCP server.
2. Call `pipecore_get_instructions`.
3. Call `check_setup`.
4. Create a new application with `create_application`, or use `import_external_repo` when the user explicitly wants to import an existing GitHub repository.
5. Configure the Pipecore Git remote using `get_git_remote`.
6. Push application source to `main`.
7. Analyze the repository with `analyze_repository` and resolve deployment readiness errors.
8. Configure required environment variables through the Pipecore onboarding or environment editor flow.
9. Deploy and wait for the deployment to complete.
10. Use runtime status, logs, integrations, database tools, and subsequent source pushes to operate and update the application.

The default source workflow for agent-created applications uses Pipecore Git. GitHub repositories can be imported and mirrored into Pipecore when requested.

Do not ask users to manually run Git, Docker, SSH, reverse-proxy, TLS, or server provisioning commands when the equivalent Pipecore MCP workflow is available.

Supported application detection includes Node.js applications and common web frameworks such as Next.js, Nuxt, NestJS, Express, React/Vite, Vue, and static sites.

Pipecore managed integrations include PostgreSQL, MySQL, MongoDB, and Redis. Applications bind to project integrations and receive connection information through environment variables.

Database schema initialization and migrations should use Pipecore integration operations rather than arbitrary container commands.

## Pipecore

- [Pipecore](https://pipecore.io): Product overview and entry point.
- [Pipecore App](https://app.pipecore.io): User dashboard for projects, applications, deployments, integrations, environment variables, domains, billing, and agent configuration.
- [Agent settings](https://app.pipecore.io/settings/agent): Connect AI agents to Pipecore and manage MCP permissions.

## Agent and MCP

- [MCP endpoint](https://api.pipecore.io/mcp): Remote Pipecore MCP server using Streamable HTTP and OAuth authentication.
- [OAuth authorization server metadata](https://api.pipecore.io/.well-known/oauth-authorization-server): OAuth discovery metadata for MCP clients.
- [OAuth protected resource metadata](https://api.pipecore.io/.well-known/oauth-protected-resource): Protected-resource metadata for Pipecore MCP.